2024 2023 2022 2020 2019 2017 2016 2015 2014 2013 2012 2011 2010 

Graduate Seminar - Melissa Dale

Author: Melissa Dale
Event Date: 2024-03-05
Location: EB 2250
Zoom: https://msu.zoom.us/j/99616578029 passcode:999341

Title: Deep Architecture Output Fusion for Domain Transfer


Abstract: This presentation outlines a model fusion technique that incorporates pre-trained deep learning models without necessitating manual fine-tuning or extensive domain knowledge. This method aims to streamline the development of efficient models, thereby broadening the accessibility of advanced machine learning methods. Our evaluation on various datasets across different fields, such as computer vision and medical imaging, confirms the technique's capability to maintain effectiveness. The findings indicate that it is feasible to achieve noteworthy accuracy without deep domain expertise or the need for significant fine-tuning. This approach facilitates easier model development, enabling both specialists and novices to utilize cutting-edge techniques for complex problem-solving. It also plays a role in making machine learning more inclusive by lowering the hurdles for entry and supporting the broader application of sophisticated models in diverse areas.